feat: Add DTS:X audio support for DASH and HLS-fMP4

packager/media/base/audio_stream_info.cc

@@ -148,6 +148,8 @@ std::string AudioStreamInfo::GetCodecString(Codec codec,
       return "dts-";
     case kCodecDTSP:
       return "dts+";
+    case kCodecDTSX:
+      return "dtsx";
     case kCodecEAC3:
       return "ec-3";
     case kCodecAC4:

packager/media/base/fourccs.h

@@ -57,6 +57,7 @@ enum FourCC : uint32_t {
   FOURCC_dtsl = 0x6474736c,
   FOURCC_dtsm = 0x6474732d,  // "dts-"
   FOURCC_dtsp = 0x6474732b,  // "dts+"
+  FOURCC_dtsx = 0x64747378,  // "dtsx"
   FOURCC_dvcC = 0x64766343,
   FOURCC_dvh1 = 0x64766831,
   FOURCC_dvhe = 0x64766865,
@@ -150,6 +151,7 @@ enum FourCC : uint32_t {
   FOURCC_trex = 0x74726578,
   FOURCC_trun = 0x7472756e,
   FOURCC_udta = 0x75647461,
+  FOURCC_udts = 0x75647473, // "udts"
   FOURCC_url = 0x75726c20,  // "url "
   FOURCC_urn = 0x75726e20,  // "urn "
   FOURCC_uuid = 0x75756964,

packager/media/base/stream_info.h

@@ -49,6 +49,7 @@ enum Codec {
   kCodecDTSL,
   kCodecDTSM,
   kCodecDTSP,
+  kCodecDTSX,
   kCodecEAC3,
   kCodecFlac,
   kCodecOpus,

packager/media/codecs/codecs.gyp

@@ -27,6 +27,8 @@
         'decoder_configuration_record.h',
         'dovi_decoder_configuration_record.cc',
         'dovi_decoder_configuration_record.h',
+        'dts_audio_specific_config.cc',
+        'dts_audio_specific_config.h',
         'ec3_audio_util.cc',
         'ec3_audio_util.h',
         'ac4_audio_util.cc',

packager/media/codecs/dts_audio_specific_config.cc

@@ -0,0 +1,29 @@
+// Copyright (c) 2023 Xperi Inc. All Rights Reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#include "packager/media/codecs/dts_audio_specific_config.h"
+
+
+#include "packager/base/logging.h"
+#include "packager/media/base/bit_reader.h"
+#include "packager/media/base/rcheck.h"
+
+namespace shaka {
+namespace media {
+
+bool GetDTSXChannelMask(const std::vector<uint8_t> &udts, uint32_t &mask) {
+  // udts is the DTS-UHD Specific Box: ETSI TS 103 491 V1.2.1 Table B-2
+  // DecoderProfileCode(6 bits)
+  // FrameDurationCode(2 bits)
+  // MaxPayloadCode(3 bits)
+  // NumPresentationsCode(5 bits)
+  // ChannelMask (32 bits)
+  BitReader bit_reader(udts.data(), udts.size());
+  RCHECK(bit_reader.SkipBits(16));
+  RCHECK(bit_reader.ReadBits(32, &mask));
+  return true;
+}
+
+}  // namespace media
+}  // namespace shaka

packager/media/codecs/dts_audio_specific_config.h

@@ -0,0 +1,23 @@
+// Copyright (c) 2023 Xperi Inc. All Rights Reserved.
+// Use of this source code is governed by a BSD-style license that can be
+// found in the LICENSE file.
+
+#ifndef PACKAGER_MEDIA_CODECS_DTS_AUDIO_SPECIFIC_CONFIG_H_
+#define PACKAGER_MEDIA_CODECS_DTS_AUDIO_SPECIFIC_CONFIG_H_
+
+#include <stddef.h>
+#include <stdint.h>
+
+#include <vector>
+
+namespace shaka {
+namespace media {
+
+class BitReader;
+
+bool GetDTSXChannelMask(const std::vector<uint8_t> &udts, uint32_t &mask);
+
+}  // namespace media
+}  // namespace shaka
+
+#endif  // PACKAGER_MEDIA_CODECS_DTS_AUDIO_SPECIFIC_CONFIG_H_

packager/media/event/muxer_listener_internal.cc

@@ -19,6 +19,7 @@
 #include "packager/media/base/video_stream_info.h"
 #include "packager/media/codecs/ec3_audio_util.h"
 #include "packager/media/codecs/ac4_audio_util.h"
+#include "packager/media/codecs/dts_audio_specific_config.h"
 #include "packager/mpd/base/media_info.pb.h"
 
 using ::google::protobuf::util::MessageDifferencer;
@@ -162,6 +163,16 @@ void AddAudioInfo(const AudioStreamInfo* audio_stream_info,
     codec_data->set_ac4_ims_flag(ac4_ims_flag);
     codec_data->set_ac4_cbi_flag(ac4_cbi_flag);
   }
+
+  if (audio_stream_info->codec() == kCodecDTSX) {
+    auto* codec_data = audio_info->mutable_codec_specific_data();
+    uint32_t channel_mask;
+    if (!GetDTSXChannelMask(codec_config, channel_mask)) {
+      LOG(ERROR) << "Failed to parse DTSX channel mask.";
+      return;
+    }
+    codec_data->set_channel_mask(channel_mask);
+  }
 }
 
 void AddTextInfo(const TextStreamInfo& text_stream_info,

packager/media/formats/mp4/box_definitions.cc

@@ -1807,6 +1807,27 @@ size_t DTSSpecific::ComputeSizeInternal() {
          sizeof(kDdtsExtraData);
 }
 
+UDTSSpecific::UDTSSpecific() = default;
+UDTSSpecific::~UDTSSpecific() = default;
+
+FourCC UDTSSpecific::BoxType() const {
+  return FOURCC_udts;
+}
+
+bool UDTSSpecific::ReadWriteInternal(BoxBuffer* buffer) {
+  RCHECK(ReadWriteHeaderInternal(buffer) &&
+         buffer->ReadWriteVector(
+             &data, buffer->Reading() ? buffer->BytesLeft() : data.size()));
+  return true;
+}
+
+size_t UDTSSpecific::ComputeSizeInternal() {
+  // This box is optional. Skip it if not initialized.
+  if (data.empty())
+    return 0;
+  return HeaderSize() + data.size();
+}
+
 AC3Specific::AC3Specific() = default;
 AC3Specific::~AC3Specific() = default;
 
@@ -1979,6 +2000,7 @@ bool AudioSampleEntry::ReadWriteInternal(BoxBuffer* buffer) {
 
   RCHECK(buffer->TryReadWriteChild(&esds));
   RCHECK(buffer->TryReadWriteChild(&ddts));
+  RCHECK(buffer->TryReadWriteChild(&udts));
   RCHECK(buffer->TryReadWriteChild(&dac3));
   RCHECK(buffer->TryReadWriteChild(&dec3));
   RCHECK(buffer->TryReadWriteChild(&dac4));
@@ -2010,7 +2032,7 @@ size_t AudioSampleEntry::ComputeSizeInternal() {
          sizeof(samplesize) + sizeof(samplerate) + sinf.ComputeSize() +
          esds.ComputeSize() + ddts.ComputeSize() + dac3.ComputeSize() +
          dec3.ComputeSize() + dops.ComputeSize() + dfla.ComputeSize() +
-         dac4.ComputeSize() + mhac.ComputeSize() +
+         dac4.ComputeSize() + mhac.ComputeSize() + udts.ComputeSize() +
          // Reserved and predefined bytes.
          6 + 8 +  // 6 + 8 bytes reserved.
          4;       // 4 bytes predefined.

packager/media/formats/mp4/box_definitions.h

@@ -334,6 +334,12 @@ struct DTSSpecific : Box {
   std::vector<uint8_t> extra_data;
 };
 
+struct UDTSSpecific : Box {
+  DECLARE_BOX_METHODS(UDTSSpecific);
+
+  std::vector<uint8_t> data;
+};
+
 struct AC3Specific : Box {
   DECLARE_BOX_METHODS(AC3Specific);
 
@@ -396,6 +402,7 @@ struct AudioSampleEntry : Box {
 
   ElementaryStreamDescriptor esds;
   DTSSpecific ddts;
+  UDTSSpecific udts;
   AC3Specific dac3;
   EC3Specific dec3;
   AC4Specific dac4;

packager/media/formats/mp4/mp4_media_parser.cc

@@ -88,6 +88,8 @@ Codec FourCCToCodec(FourCC fourcc) {
       return kCodecDTSL;
     case FOURCC_dtse:
       return kCodecDTSE;
+    case FOURCC_dtsx:
+      return kCodecDTSX;
     case FOURCC_dtsp:
       return kCodecDTSP;
     case FOURCC_dtsm:
@@ -489,6 +491,9 @@ bool MP4MediaParser::ParseMoov(BoxReader* reader) {
           max_bitrate = entry.ddts.max_bitrate;
           avg_bitrate = entry.ddts.avg_bitrate;
           break;
+        case FOURCC_dtsx:
+          codec_config = entry.udts.data;
+          break;
         case FOURCC_ac_3:
           codec_config = entry.dac3.data;
           num_channels = static_cast<uint8_t>(GetAc3NumChannels(codec_config));

packager/media/formats/mp4/mp4_muxer.cc

@@ -83,6 +83,8 @@ FourCC CodecToFourCC(Codec codec, H26xStreamFormat h26x_stream_format) {
       return FOURCC_dtse;
     case kCodecDTSM:
       return FOURCC_dtsm;
+    case kCodecDTSX:
+      return FOURCC_dtsx;
     case kCodecEAC3:
       return FOURCC_ec_3;
     case kCodecAC4:
@@ -493,6 +495,9 @@ bool MP4Muxer::GenerateAudioTrak(const AudioStreamInfo* audio_info,
       audio.ddts.sampling_frequency = audio_info->sampling_frequency();
       audio.ddts.pcm_sample_depth = audio_info->sample_bits();
       break;
+    case kCodecDTSX:
+      audio.udts.data = audio_info->codec_config();
+      break;
     case kCodecAC3:
       audio.dac3.data = audio_info->codec_config();
       break;

packager/mpd/base/xml/xml_node.cc

@@ -43,6 +43,9 @@ typedef MediaInfo::VideoInfo VideoInfo;
 namespace {
 const char kEC3Codec[] = "ec-3";
 const char kAC4Codec[] = "ac-4";
+const char kDTSCCodec[] = "dtsc";
+const char kDTSECodec[] = "dtse";
+const char kDTSXCodec[] = "dtsx";
 
 std::string RangeToString(const Range& range) {
   return base::Uint64ToString(range.begin()) + "-" +
@@ -606,6 +609,19 @@ bool RepresentationXmlNode::AddAudioChannelInfo(const AudioInfo& audio_info) {
                            "1");
     }
     return ret;
+  } else if (audio_info.codec() == kDTSCCodec ||
+             audio_info.codec() == kDTSECodec) {
+    audio_channel_config_value = base::UintToString(audio_info.num_channels());
+    audio_channel_config_scheme =
+      "tag:dts.com,2014:dash:audio_channel_configuration:2012";
+  } else if (audio_info.codec() == kDTSXCodec) {
+    const auto& codec_data = audio_info.codec_specific_data();
+    const uint32_t channel_config =
+      base::HostToNet32(codec_data.channel_mask());
+    audio_channel_config_value =
+      base::HexEncode(&channel_config, sizeof(channel_config));
+    audio_channel_config_scheme =
+      "tag:dts.com,2018:uhd:audio_channel_configuration";
   } else {
     audio_channel_config_value = base::UintToString(audio_info.num_channels());
     audio_channel_config_scheme =